Staurolite and Other Aluminous Phases in Alpine Eclogite from the Central Swiss Alps: Analysis of Domain Evolution

105

The Canadian Mineralogist Vol. 43, pp. 105-128 (2005)

STAUROLITE AND OTHER ALUMINOUS PHASES IN ALPINE ECLOGITE FROM THE CENTRAL SWISS ALPS: ANALYSIS OF DOMAIN EVOLUTION

FRAUKJE M. BROUWER§ AND MARTIN ENGI§ Institute of Geological Sciences, University of Bern, Baltzerstrasse 1, CH–3012 Bern, Switzerland

ABSTRACT

Kyanite eclogite in the Central Swiss Alps, partially hydrated during decompression, shows conspicuous domains containing unusual phases, such as staurolite, corundum, and hercynite-dominant spinel. Detailed analysis of these domains indicates that they formed by prograde growth of porphyroblasts of lawsonite, kyanite, and garnet, prior to the complete conversion of igneous plagioclase to omphacite. In the evolution following the culmination in pressure, the chemically differentiated domains remained largely distinct from the matrix. Quantitative thermodynamic models set up for each domain allow us to analyze each domain type, in order to test (a) to what extent they behaved as closed systems, (b) whether local equilibrium was maintained during decompression, and (c) what reaction mechanisms and volumes were involved in the observed transformations. Results indicate that metasomatic effects were quite limited; local equilibration was possible to mid-crustal levels at temperatures near 700°C; reaction paths followed involved minimal transport of Al. Both prograde and retrograde parts of the P–T path could be deduced for eclogite samples from the Southern Steep Belt of the Central Swiss Alps. A late-orogenic, prograde thermal overprint (heating spike) is confirmed.

Keywords: reaction volume, kyanite eclogite, decompression, thermodynamic modeling, local equilibrium, DOMINO program, Swiss Alps.

SOMMAIRE

Les éclogites à kyanite des Alpes suisses centrales, partiellement hydratées au cours de la décompression, font preuve de domaines bien visibles contenant des minéraux inhabituels, par exemple staurolite, corindon, et un spinelle à dominance de hercynite. Une analyse détaillée de ces domaines montre qu’ils se sont formés par croissance prograde de porphyroblastes de lawsonite, kyanite, et grenat, avant la conversion complète du plagioclase igné en omphacite. Au cours de l’évolution suivant le paroxysme en termes de pression, les domaines chimiquement différenciés sont restés distincts de la matrice. Des modèles thermodynamiques quantitatifs établis pour chaque domaine nous permettent d’analyser chaque type de domaine afin de savoir (a) à quel point ils se sont comportés en systèmes fermés, (b) si l’équilibre local a été maintenu au cours de la décompression, et (c) quels mécanismes et volumes de réaction ont été impliqués dans les transformations observées. Les résultats indiquent que les effets métasomatiques ont été assez limités; l’équilibrage local pendant la décompression a été possible à un niveau équivalent au milieu de la croûte, à des températures d’environ 700°C, et les séquences de réaction ont été régies par la distance minimale de transfert de l’aluminium. On peut en déduire les tracés prograde et rétrograde de l’évolution P–T des échantillons d’éclogite provenant du “Southern Steep Belt” des Alpes centrales suisses. L’importance d’un événement thermique prograde tardi- orogénique (pointe thermique surimposée) est confirmée.

(Traduit par la Rédaction)

Mots-clés: volume de réaction, éclogite à kyanite, décompression, modèle thermodynamique, équilibre local, logiciel DOMINO, Alpes suisses.

§ E-mail address: [email protected], [email protected] 106 THE CANADIAN MINERALOGIST

INTRODUCTION ever, such rocks may provide information on petroge- netic conditions and processes at several stages of their Eclogites from various places in the world have been evolution, or on how the domains formed in the first reported to contain symplectites of plagioclase and Al- place, and to what extent they may be regarded as closed rich minerals such as corundum, Fe–Mg-bearing spinel systems subsequently. We aim to show that under cer- and sapphirine surrounding relics of kyanite or zoisite tain conditions, useful models can be developed that (e.g., Dal Vesco 1953, Lappin 1960, Liati & Seidel allow an assessment of the actual volumes of equilibra- 1994). Staurolite has been reported as a product of section, an interpretation of the texture, inferences on likely ondary replacement in eclogites from eastern China precursor phases, and an estimate of the P–T conditions (Enami & Zang 1988) and Greenland (Elvevold & at which hercynite-dominant spinel, corundum and stau- Gilotti 2000). As several investigators have noted, such rolite may be stabilized in eclogites. aluminous phases would not be expected in rocks of basaltic bulk composition if these rocks behaved as a METHODS single equilibration volume. The development of compositionally differentiated domains during metamor- This study relies on a comparison of stable phase- phism may indicate spatial limits to thermodynamic assemblages computed for a specified bulk-composition equilibrium and implies at least grain-scale mass trans- with those documented in kyanite eclogite samples. fer. Disequilibrium considerations have made some au- With the programs DOMINO and THERIAK (De Capitani thors reluctant to apply equilibrium models in order to 1994), we calculate the complete, stable assemblage of draw conclusions on the conditions of formation of minerals for each point in a P–T grid for a given bulk- symplectites (e.g., Elvevold & Gilotti 2000, Sabau et composition. These programs can be used to constrain al. 2002). Yet, on the basis of an assumption of local the conditions at which a given assemblage of minerals chemical equilibrium, an elegant kinetic model was re- equilibrated, as well as possible precursor assemblages. cently developed for the decompression of kyanite Inversely, where the models fail to reproduce the ob- eclogites (Nakamura 2002). It showed how chemical served assemblage, inferences can be made regarding potential gradients of Na2O and CaO due to garnet and the accuracy of the estimated bulk-composition, gains clinopyroxene breakdown cause the formation of a pla- or losses of components from the domain, and the pos- gioclase mantle between kyanite and quartz. Local SiO2 sibility of disequilibrium among the phases present undersaturation within this mantle allows kyanite to within the domain. survive decompression. Even in the presence of an aque- Accurate modeling of the evolution of assemblages ous fluid, such undersaturation prevents reactions with of metamorphic minerals is possible only where the garnet, clinopyroxene and quartz, which would other- chemical composition within a reaction volume (or wise produce amphibole and plagioclase. In this SiO2- equilibration volume) is known. Conventional bulk- deficient area, corundum and Fe–Mg-bearing spinel compositions are not appropriate in this study because may develop, although free quartz is ubiquitous outside mm-size domains, containing quartz-undersaturated the plagioclase mantle. metamorphic assemblages, occur in a matrix with free The current study of symplectites from the Central quartz. Effective reaction-volumes thus probably were Alps is also based on an assumption of local equilib- of the same magnitude as the domains observed. Where rium. Using careful estimates of the “bulk” chemical possible, we selected domains in a thin section such that composition of the various symplectite domains ob- they correspond roughly to a central section. We then served, we investigate to what extent equilibrium ther- assumed that the total surface-area of each mineral re- modynamic models can predict assemblages of minerals flects its modal abundance and converted these modes observed in kyanite eclogite symplectites, notably those to an effective bulk-composition using averaged results leading to the appearance of corundum, hercynite-domi- of electron-microprobe analyses. This procedure is a nant spinel, and staurolite. One goal of the study is to rough approximation at best, for at least two reasons: analyze whether and under which conditions it is pos- (a) for domains of ovoid or spheroid shape, the contri- sible to model particular mineralogical and textural do- bution of the rim composition would be underestimated; mains using a strictly equilibrium thermodynamic (b) for any shape, a section far from the center of a do- approach. We realize that although the geometry of co- main would underestimate the contribution of the core. rona structures and symplectites provides petrographic In the absence of precise data for (b) from either tomog- evidence of limited chemical interaction with their raphy or systematic sectioning, and in view of the pris- chemically different surroundings, it is not trivial to use matic or lozenge-shaped domains analyzed here, it these observations to infer reaction volumes. Retro- seems difficult to improve upon the simple assumption graded eclogites may have complex pressure – tempera- made. ture – deformation (P–T–D) paths, and the chemical We reconstruct the local bulk-composition for each contents, mobility and episodic interaction with fluids type of domain by analysis of back-scattered electron are certain to have varied along the prograde and retro- images and, where necessary, use of X-ray maps (both grade metamorphic paths followed by these rocks. How- on Camscan CS4 SEM, University of Bern). Electron- DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 107 microprobe analyses were performed using a Cameca show evidence of subsequent Barrovian high-tempera- SX–50 (15 kV and 20 nA) at the University of Bern. ture metamorphism at intermediate pressure, which Synthetic and natural mineral standards were used, and caused extensive migmatization in the felsic and inter- PAP corrections applied. mediate lithologies enveloping the mafic lenses. In modeling the domain, modal abundances and nor- The SSB is considered to represent an exhumed por- malized compositions of minerals are selected to calcu- tion of a tectonic accretion channel (TAC, Engi et al. late a preliminary bulk-composition as input for DOMINO. 2001) related to the Tertiary collision of Eurasia with This approach implies a certain amount of judgment the Apulian promontory of Africa (Schmid et al. 1989, regarding the delimitation of any domain against its 1996). This TAC is a mélange zone 5–8 km in thick- neighbors, as well as the selection of specific analytical ness, which formed at the contact of the subducting slab data (or means thereof) for minerals contained within and the overriding plate. The TAC accommodates tec- the domain. Out of the four examples presented below, tonic fragments considered to be derived from both preliminary calculations using THERIAK for two cases (A, plates during the subduction, accretion, and exhumation. B) showed that the initial choices of domain boundaries We note that in the context of such a TAC, it is expected led to predicted assemblages of minerals that reproduced that lenses may record different P–T trajectories within the observed assemblages and compositions of miner- the same thermal-tectonic regime, and that the exhumed als. In the two other cases (domains C and D), it was channel behaved as a tectonic mélange, rather than a not evident how much quartz was effectively part of the coherent nappe, at least during initial parts of the his- domains, because they are surrounded by free quartz. tory. In the Adula nappe, which represents a part of this Our initially estimated bulk-compositions included 6 same mélange unit (partly outside the SSB), the partial and 5 vol.% of quartz, respectively, but preliminary cal- hydration and amphibolite-facies overprint of eclogite culations showed that this additional SiO2 had to be re- lenses upon decompression have long been established moved for the stable assemblages of minerals predicted (Heinrich 1986). to approach the observations closely. The samples selected for the current study were col- An essential condition for the success of our ap- lected from two lenses of kyanite eclogite at Gorduno proach is the availability of adequate thermodynamic (indicated in Fig. 1b), near Bellinzona in the Central models, including solution models, for all phases with a Swiss Alps (Grubenmann 1908, Möckel 1969). potential stability-field in the P–T region of interest. The Eclogites from the Gorduno area have a tholeiitic bulk- thermodynamic database used in this study is an ex- composition (Bocchio et al. 1985). The lenses are tended version of the database of Berman (1988), with mostly made up of biotite and garnet amphibolites, but internally consistent additions by Meyre et al. (1997) occasional relics of kyanite and very rarely omphacite for omphacite (diopside, jadeite and hedenbergite end- are preserved. Conditions of high-pressure metamor- members), Meyre et al. (1999) for annite, glaucophane, phism were in the range 750 ± 50°C at 23 ± 3 kbar (Tóth ferroceladonite and celadonite, and Nagel (2002) for et al. 2000). Relics of the high-pressure assemblage are staurolite and Fe-dominant chlorite. Also added were usually surrounded by symplectites of hornblende and an ideal speciation model for hercynite-dominant spinel plagioclase around garnet, and of amphibole ± (based on Engi 1983), a regular asymmetric model for clinopyroxene + plagioclase replacing omphacite. The cummingtonite–grunerite amphiboles (based on data rare occurrence of spinel and corundum in symplectites from Ghiorso et al. 1995), and a new site-occupancy around kyanite was previously noted (Dal Vesco 1953, model for Fe–Mg-bearing hornblende, based on Mäder Möckel 1969, Bocchio et al. 1985), and recently a first & Berman (1992) for the pseudobinary join actinolite – attempt at retrieving information about the conditions pargasite. The latter two models are but first-order ap- of their formation yielded an estimate of 750 ± 40°C at proximations to describe the amphiboles occurring as 8 ± 1 kbar (Tóth et al. 2000). Figure 2 is an equilibrium minor phases in some of the coronas. A complete copy phase-diagram for an average bulk-composition of of the thermodynamic database used is available from Gorduno eclogite, which we consider representative of the Depository of Unpublished Data, CISTI, National the samples presented in this study (Table 2, adapted Research Council, Ottawa, Ontario K1A 0S2, Canada, from Bocchio et al. 1985), with hydrous fluid added. It or from the authors on request. indicates the typical parageneses of minerals one would expect for a fully equilibrated eclogite, hydrated during GEOLOGICAL CONTEXT AND SAMPLE DESCRIPTION decompression. At the conditions of the Barrovian overprint, the predicted assemblage is dominated by horn- The Southern Steep Belt (SSB) in the Central Alps blende and intermediate plagioclase, with additional of southern Switzerland (Fig. 1a) is largely made up of cummingtonite (10 vol.%), omphacite (8%), ilmenite a composite of granitic gneisses, metaclastic schists, (1.7%) and H2O. It is notable that none of the alumi- marbles and calc-silicate rocks that incorporates centi- nous phases mentioned above are predicted to occur for meter- to kilometer-scale lenses of mafic and ultrama- this bulk composition. fic material. Some rock types in this high-strain zone In the present study, we focus on four types of coro- preserve relics of high-pressure metamorphism, and all nas from three samples of garnet amphibolites highly 108 THE CANADIAN MINERALOGIST affected by symplectite formation. Garnet in these rocks kyanite, hercynite-dominant spinel, corundum and stau- is invariably surrounded by a rim of amphibole and pla- rolite, in which amphibole is very rare. Figure 3 depicts gioclase, locally with secondary (low-Na) clino- a typical sample in hand specimen, in which the size pyroxene. Very few relics of primary omphacite remain; and distribution of different corona domains are quite most of it was replaced by very fine symplectitic visible. Four examples of such coronas are investigated intergrowths of secondary clinopyroxene, amphibole in detail below: domain A: high-An plagioclase with and plagioclase. The samples contain symplectites hercynite and corundum, domain B: kyanite and stauro- dominated by plagioclase and aluminous phases like lite, domain C: garnet and corundum, and domain D:

FIG. 1. (a) Tectonic map of the Central Swiss Alps. The Southern Steep Belt is the region where exhumed units are steeply oriented along the Insubric Line, marked in light red. Location of the map shown in (b) is indicated. After Pfiffner & Trommsdorff (1998). (b) Geological map indicating the locations of the kyanite eclogite boudins from which the samples studied were collected. After Bächlin et al. (1974). DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 109

FIG. 2. TCNFMASH equilibrium phase-diagram computed with DOMINO for average Gorduno eclogite (after Bocchio et al. 1985; Table 2), with hydrous fluid added. The P– T path is for retrograded eclogites from Gorduno (Tóth et al. 2000). The predicted fully equilibrated hydrous amphibolite-facies assemblage is listed for 650°C, 5.5 kbar (star), the conditions for the regional Barrovian overprint. The compositions of the phases at that stage, computed using THERIAK, are given in Table 3. Reaction labels indicate the appearence of new phases; shaded fields denote the stability of garnet (light grey) and garnet + rutile (dark grey). Upon decompression, at around 10 kbar, rutile is predicted to disappear from the eclogite assemblage, followed by garnet at around 6 kbar. Abbre- viations mostly follow Kretz (1983); all are listed in Appendix 1.

FIG. 3. Polished slab of sample Ma9338 showing garnet relics (pink) in a matrix dominated by hornblende (black) and plagioclase (white). Garnet is rimmed by hornblende + plagioclase symplectite. The bright corona near the center of the image is domain A, consisting of plagioclase with tiny inclusions of corundum and spinel. Domain A is discussed in detail below. 110 THE CANADIAN MINERALOGIST kyanite and hercynite. Samples AA97–16.1 (with do- DOMAIN A: HIGH-AN PLAGIOCLASE main B) and AA97–14.2 (domains C and D) were col- WITH HERCYNITE AND CORUNDUM lected from the same eclogite boudin (Loc. 2 in Fig. 1b, Swiss coordinates 721.8/120.0/690m). Although they In this and the following sections, we summarize have experienced the same pressure–temperature evo- results of our petrographic and microchemical study of lution, they may well have recorded different stages four types of domains observed in the three samples of along this trajectory. Since sample Ma9338 (with do- kyanite eclogite described above. These data are pre- main A) was collected from a different boudin (Loc. 1 sented in comparison to the thermodynamic models and in Fig. 1b, 722.7/119.9/390m) within the mélange se- inferences therefrom regarding possible precursor quence, it may have gone through a significantly differ- phases in each domain. ent P–T history. A precursor study by Tóth et al. (2000) Appendix 2 lists averaged results of electron-micro- concentrated on garnet domains at this same locality; probe analyses of the phases found in each domain (A – therefore, this type of domain is not investigated further D). Table 1 contains modal abundances of all phases in here. these and, in the case of solid solutions, their mineral compositions; in addition, the preliminary estimates (A’ – D’) of local bulk-compositions computed from the above data are given. Table 2 lists the simplified bulk compositions used as input for THERIAK and DOMINO, as well as the slightly adjusted input compositions (A”, C”, and D”) used in further modeling. Table 3 contains the modal abundances and compositions of minerals that define stable assemblages calculated for specific points, indicated with stars in the corresponding P–T diagrams. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 111

Characterization of domain A An30–50 (Fig. 4c), which includes one large grain of apatite (~200 m) and rutile (60 m) each. The inner por- The overall corona in sample Ma9338a is oval in tion of the corona is texturally different, with much more shape, slightly larger than 1 2 mm, and is made up of conspicuous grain-boundaries and ubiquitous fine inclu- mostly plagioclase. It is surrounded by a mixture of cal- sions. The shape of this zone is reminiscent of lawsonite cic amphibole with fine plagioclase (Figs. 4a, b). The or zoisite (Fig. 4b), and it is made up of mostly plagio- lower edge of the domain abuts against two smaller clase, with a composition of An78–98 (mean An90). The garnet domains, rimmed by plagioclase and amphibole, inner zone contains tiny inclusions of ilmenite, as well which impinge on the outer rim of plagioclase in that as a phase that is so fine-grained we have been unable section. The domain has an outer rim of plagioclase with to analyze it well. “Mixed-target” microprobe data com-

The amount of free H2O present as steam is given in moles. The symbols *, + and o after the figure numbers correspond to symbols in the P-T diagrams. 112 THE CANADIAN MINERALOGIST

FIG. 4. Domain A. (a) Thin section photograph in plane-polarized light. (b) Back-scattered SEM image with indexed colors. (c) Variation of plagioclase components in electron- microprobe analyses along the transect indicated in Figure 4b. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 113 monly yield low totals. Only Si, Al, Ca, Na and K (up occur, albeit again at very low modal abundance (~1 to 7.7 wt% K2O) were detected, and the SiO2 content is vol.%; Table 3). Garnet is missing in the present sec- consistently higher than for plagioclase (but always <58 tion through the domain, but its presence in other sec- wt%). Based on these findings, K-feldspar would seem tions through this domain cannot be excluded. a likely candidate. As the compositions plot in the “for- Because K was left out of the simplified bulk-com- bidden zone” of the feldspar diagram, plagioclase and position, muscovite and K-feldspar do not appear in the K-feldspar would have coexisted at some stage. How- computed equilibrium phase-diagram. Muscovite flakes ever, micro-Raman spectra of the potassic phase corre- are associated with a crack, and we believe that musco- spond neither to the spectrum of K-feldspar nor that of vite formed in response to an influx of aqueous fluid. muscovite, phlogopite or leucite. Because it has not been possible to identify the potassic phase, and the total Precursor amount of K in the domain is but 0.25 wt% K2O (Table 1), we chose to omit that component from fur- As noted above, the shape of the present domain is ther analysis. The inner zone contains a few grains of reminiscent of lawsonite or zoisite, and the predomi- corundum, as well as strings of hercynite (Hc71Spl29). nance of bytownite indicates that these minerals are In the core of the inner zone, a few grains of pure mus- indeed likely precursors for the domain. These possi- covite occur, which are roughly lined up along a crack, bilities can be assessed in a binary phase-diagram, com- whereas the outside includes some grains of puted for a simplified bulk-composition A” (Table 2) magnesiohornblende. with variable amounts of H2O versus temperature at a The inner zone of high-An plagioclase (light pink in pressure within the stability fields of lawsonite and Fig. 4b) with its inclusions is called domain A and re- zoisite, arbitrarily set at 14 kbar. In comparison with garded as a remnant of some porphyroblast, which was not visibly affected by diffusional equilibration with the matrix. (In contrast, the outer rim is attributed to such interaction.) It is noteworthy that the range of compositions in the inner domain suggests plagioclase straddling the Huttenlocher gap (Carpenter 1985), whereas the outer zone may reflect the Bøggild gap for temperatures of ~600°C. Neither apatite nor rutile is considered part of domain A. Ti and K (as discussed above) are present in very limited amounts; hence they were left out of the simplified bulk-composition of the domain (A’ in Table 2). Figure 5 shows an ACF plot of the bulk composition of the eclogite sample and the plagioclase-rich domain (as well as the domains presented below) and thus depicts some of the effect of metamorphic differentiation in the sample.

Model for the retrograde assemblage

The equilibrium phase-diagram computed with the FIG. 5. ACF plot with the bulk compositions calculated for program DOMINO for the estimated bulk-composition A’ each of the domains, as well as the average bulk-composi- is depicted in Figure 6. Owing to the absence of quartz, tion of the Gorduno eclogite from Bocchio et al. (1985). the stability field of corundum is large and overlaps to a Notably, the calculated compositions of the domains plot major extent that of the Fe–Mg-bearing spinel. Isopleths in different places, all far from the eclogite whole-rock for hercynite in the Fe–Mg-bearing spinel indicate for- composition. Projection from SiO2 and H2O. Closed dia- mation temperatures of 600–700°C at pressures below monds represent quartz-saturated bulk compositions, open about 8 kbar. If the magnesiohornblende occurring in diamonds are quartz-undersaturated. Colored fields refer the rim of the domain is included in equilibrium consid- to assemblages formed during decompression reactions, the erations, the P–T conditions are constrained to 6.5 ± 1 dashed field to the predicted amphibolite-facies assemblage kbar and 675 ± 25°C (shaded field in Fig. 6). Because for a fully equilibrated hydrated eclogite. Garnet and hornblende compositions are as analyzed in the domains de- of uncertainties in the solution model for amphibole, and scribed. Minerals preserved only as relics of HP stage are owing to the low modal abundance of magnesio- indicated in brackets. Since only quartz was removed from hornblende, its exact composition cannot be used to fur- the initially calculated bulk-compositions, C” and D” plot ther constrain the equilibrium conditions. Within the in the same location as C’ and D’, respectively. A” would indicated field of stability, garnet is also predicted to plot on the A–C axis, because Fe and Mg were excluded from A”. 114 THE CANADIAN MINERALOGIST bulk composition A’, A” no longer includes FeO and breakdown of lawsonite and zoisite is assumed to have MgO, and excess H2O. For A”, zoisite, lawsonite or both escaped and thus was no longer available on the retro- will be stable under these conditions (Fig. 7a). The grade path. modal abundance of zoisite never exceeds 70%, whereas a substantial excess H2O along the prograde path will DOMAIN B: KYANITE AND STAUROLITE lead to almost 89% lawsonite of the total volume (Table 3). An equilibrium phase-diagram for bulk A” confirms Characterization of the domain that at high temperature and intermediate pressure, lawsonite breaks down to plagioclase (98%), with traces Domain B has a core of kyanite intergrown with of corundum and sillimanite (Fig. 7b, Table 3). Where staurolite (Fig. 8, Table 1). The mantle is made up of Fe and Mg are taken into account, sillimanite would be plagioclase, ranging from An78 adjacent to kyanite, to replaced by Fe–Mg-bearing spinel (Fig. 6). Pure An49 in contact with staurolite, to An26 at the outer rim lawsonite would break down to pure anorthite and va- (Figs. 8c, d). Around kyanite, a thin veneer of K-feld- por, whereas pure zoisite breaks down to anorthite spar is locally present. The edge of the plagioclase (68%), grossular (29%) and corundum (3%). In view of mantle is defined by strings of low-Ca amphibole the shape of the domain and the close compositional (Cum64Gru36), and outside is mostly quartz. Locally, a match, a lawsonite porphyroblast is the most likely pre- few grains of magnesiohornblende are present at the cursor for domain A. The formation of porphyroblasts edge of the plagioclase mantle, which appears to be a of lawsonite under equilibrium conditions sets narrow product of alteration of cummingtonite. A few grains of P–T limits (Fig. 7b) for the prograde path of evolution ilmenite represent the only accessory phase in the do- of this type of domain. Note that the fluid liberated upon main. The problem of formulating a model composition

FIG. 6. CNFMASH equilibrium phase-diagram, domain A, bulk A’. For a few fields of stability, the stable assemblage of minerals is listed, in order of modal abundance. Star and plus symbols indicate conditions for which the predicted stable assemblage, including modal abundance and mineral compositions, is listed in Table 3. The symbols do not represent a P–T path for the rocks studied. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 115

FIG. 7. (a) Binary diagram for H2O versus T with bulk A” at 14 kbar, in CNASH. (b) Inset: CNASH equilibrium phase-diagram for bulk A” (wet), indicating the stability field of lawsonite and breakdown sequence to high temperatures. Star, circle and plus symbols refer to predicted stable assemblages listed in Table 3.

for this domain (as well as C and D) stems from the cating a pressure below 9 kbar is imposed by the ab- strong compositional zoning of the plagioclase shell. A sence of hornblende and garnet from domain B. How- composition of XAn = 0.33 was selected using composi- ever, because the calculations only predict very small tional profiles (Figs. 8c, d), being representative of the modal amounts of hornblende and garnet, that limit is spread of data and the thickness of each zone. less certain, especially since these phases might be present outside the plane of the thin section. The pre- Model for the retrograde assemblage dicted modes and compositions closely correspond to those determined in domain B (Table 3). The formation The estimated bulk-composition of this domain (B’ of magnesiohornblende at the expense of cumming- in Table 2, Fig. 5) yields an equilibrium phase-diagram tonite is predicted to have occurred soon after cooling in which the stability field of the observed assemblage below the inferred conditions of equilibration (Fig. 9). of minerals is constrained by the absence of chlorite We note, however, that no chlorite formed inside the (>520°C), and the presence of staurolite (<670°C, domain, probably for lack of sufficient aqueous fluid. Fig. 9). Isopleths for the composition of staurolite suggest that it formed between 540 and 630°C. The presence of kyanite and cummingtonite indicates pressures between 4 and 10 kbar. An additional constraint indi- 116 THE CANADIAN MINERALOGIST

Precursor previous estimates from the area (e.g., Brouwer 2000, Tóth et al. 2000). If less SiO2 is included in the domain, The equilibrium assemblage predicted by the pro- the abundance of free quartz is affected, with little in- gram THERIAK for the bulk composition of domain B at fluence on the relative abundances of the other phases. 700°C and 20 kbar is dominated by 48 vol.% omphacite There is no single obvious precursor phase for this do- and 22% kyanite (Table 3). These conditions of high- main, but since kyanite porphyroblasts make up a fairly pressure metamorphism were chosen to approximate small part of the volume of kyanite eclogites, it is likely

FIG. 8. Domain B. (a) Photograph of a thin section in plane-polarized light. (b) Back-scattered SEM image with indexed colors. (c, d) Variation of plagioclase components (electron-microprobe data) along the transects indicated in Figure 8b. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 117 that this domain formed where a kyanite porphyroblast magnesiohornblende, orthopyroxene (En58Fs39) and interacted with neighboring omphacite. Material-trans- quartz near the outer edges. Plagioclase in the mantle fer analysis confirms the conclusion reached by Godard ranges from An55 at the rim of garnet to An16 on the & Mabit (1998) that omphacite must be involved in the outside (Fig. 10d). Again, the plagioclase mantle is formation of corundum, Fe–Mg-bearing spinel and sap- rimmed by quartz. A few grains of rutile are included in phirine-bearing symplectites after kyanite. garnet, whereas ilmenite is present in the rest of the domain. Locally, chlorite occurs as a retrograde replace- DOMAIN C: GARNET AND CORUNDUM ment of phlogopite.

Characterization of the domain Model for the retrograde assemblage

Domains C and D are neighbors in sample AA97– The initial models for this domain indicated that only 14.2 (Figs. 10a, b). In the core of domain C (lower left little quartz may be included in the reconstruction of the in Fig. 10b), a grain of garnet and many grains of co- bulk composition of the domain, because otherwise up rundum are distributed among phlogopite grains in a to 7 vol.% of kyanite or sillimanite would be predicted, matrix of plagioclase (An36–54). The garnet has a fairly whereas neither phase is present within the domain. In flat zoning pattern, with Alm52Grs25Prp21Sps2 in a broad addition, small amounts of quartz produce a relatively core zone, and a 30-m rim with the composition grad- large joint field of stability for garnet, corundum and ing to Alm57Grs19Prp18Sps6 at the edge. The mantle biotite (Fig. 11, for quartz-free bulk C”, Table 2). To consists mostly of plagioclase, with inclusions of

FIG. 9. Calculated CNFMASH equilibrium phase-diagram delimiting the stability fields of all mineral assemblages for domain B with bulk composition B’ (Table 2). Isopleths for Mg in staurolite. The predicted stable assemblage of minerals in the shaded field is listed in order of modal abundance. Star symbol refers to predicted stable assemblage listed in Table 3. 118 THE CANADIAN MINERALOGIST reach an adequate match of prediction and observation, DOMAIN D: KYANITE AND HERCYNITE the bulk composition was adjusted slightly from the initial estimate (C’ in Table 2, Fig. 5, with 6 vol.% quartz Characterization of the domain included). Our preferred model for the current domain predicts Domain D (top right in Fig. 10b) is dominated by a equilibration temperatures above 700°C owing to the core of kyanite, surrounded by fine hercynite absence of staurolite (Fig. 11). The pressure is con- (Hc75Spl25) and coarser phlogopite (XMg in the range strained between 5 and 11 kbar by the absence of white 0.59–0.66). These minerals are embedded in plagio- mica and Fe–Mg-bearing spinel from the domain (Figs. clase, with a composition of An52–63 between the kyan- 10, 11). These constraints are considered firm, because ite grains, and ranging from An88 bordering kyanite to in garnet, biotite and corundum, all three constraining An17 at the outside edge (Fig. 10c). The most calcic pla- phases have a likely precursor at whose expense they gioclase also has a small orthoclase component (2%). are likely to have formed. The presence of ilmenite in- Beyond the plagioclase mantle, the domain is sur- stead of rutile in the domain matrix suggests equilib- rounded by quartz, with minor amounts of edenite and rium at 6.5 ± 1 kbar and T > 720°C. There is no upper orthopyroxene (En59Fs38) occurring at the contact be- temperature constraint in this domain, and higher tem- tween plagioclase and quartz. Some grains of ilmenite peratures would imply significantly higher pressures of are distributed among the hercynite strings. equilibration. Higher temperatures, however, are considered unlikely in view of the relatively low tempera- Model for the retrograde assemblage tures of equilibration inferred for domain B, from a different sample collected from the same eclogite If some quartz is included in the estimated bulk- boudin. composition (D’ in Table 2, Fig. 5), we compute with The model fails to reproduce the observed composi- DOMINO an equilibrium phase-diagram in which the sta- tion of garnet in the stability field of biotite and corun- bility fields of Fe–Mg-bearing spinel and kyanite are dum, but does so at significantly higher P and T (e.g., at separated by about 150°C and 4 kbar. Also, Fe–Mg- 700°C and 16 kbar, Table 3). The garnet in the domain bearing spinel is predicted to be stable with cordierite, is apparently the remaining core of a grain of garnet that which is not observed in domain D, and at temperatures was partly resorbed. We note that the composition of over 850°C, which seem exceedingly high. With no garnet and biotite cannot be used to determine exact P– quartz in the domain (D” in Table 2), the stability fields T conditions of their formation, because exchange with are much closer, but still do not overlap (Fig. 12a). hornblende, which affects garnet and biotite, cannot be One of the phlogopite samples in this domain has modeled accurately. If calculated, the isopleth for the XMg equal to 0.64, and three others cluster around 0.60. appropriate composition of biotite actually plots below If the former equilibrated with kyanite, this constrains the P–T field in Figure 11. The formation of ortho- the P–T conditions at 750 ± 25°C and 9.5 ± 1 kbar. Iso- pyroxene and magnesiohornblende in the rim of the pleths for hercynite in Fe–Mg-bearing spinel, and the domain is not easily explained with the current model. disappearance of corundum define a stability field of It is certainly possible that inadequacies of our amphi- hercynite at 720 ± 25°C and 5.0 ± 1 kbar, again inside bole solution-models are responsible for the failure to the stability field of ilmenite. A binary phase-diagram reproduce the observed phase-relations in detail. Alter- for pressure versus SiO2 contents indicates that the natively, interaction with other domains outside the field amount of quartz originally assumed to belong to this of view or the plane sectioned could account for the domain was slightly overestimated. Decompression minor discrepancy. In similar rocks, orthopyroxene has from 8 to 5 kbar at about 700°C results in hercynite been reported where garnet abuts against a symplectite growth, while avoiding the corundum field (Fig. 12b), of sapphirine and plagioclase after kyanite (Carswell et only if less than 3% (instead of the original 5%) of free al. 1989), or as a breakdown product of garnet or quartz are included. We therefore interpret the forma- clinopyroxene or both (Johansson & Möller 1986). tion of hercynite in domain D to result from the breakdown of garnet ± staurolite. Precursor The model also predicts the growth of considerable amounts of sillimanite (Table 3), which is not observed For the bulk composition of domain C at 700°C and in domain D. The transformation from kyanite to silli- 20 kbar, we computed with THERIAK a local equilibrium manite may have been hindered by the limited avail- assemblage dominated by omphacite, white mica, and ability of a fluid phase that could have enhanced the kyanite, with contributions of garnet, quartz and reaction kinetics. The biotite, with a phlogopite compo- paragonite (Table 3). Again, a likely scenario is that a nent around 0.60, as well as magnesiohornblende at the kyanite porphyroblast and its omphacite neighbors in- rim of the domain, likely have formed upon decompres- teracted to produce the present domain. sion (Fig. 12a). The observed presence of orthopyroxene DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 119 is not predicted by this model, which, as in domain C, to have been derived from crystallizing partial melts in may be due to the imperfection of our amphibole solu- the immediate vicinity, as migmatitic gneisses with up tion model or to minor chemical interaction with neigh- to 30 vol.% leucosome and occasional pegmatitic accu- boring domains. mulations envelop the mafic lenses. Hydrous fluid ex- pelled during crystallization of these granitic or tonalitic Precursor melts were certainly saturated in aqueous silica and possibly in feldspar components; hence some mass For the quartz-free bulk composition (D”) at 700°C transfer into the basic boudins is probable. A sparse and 20 kbar, we compute with THERIAK a local equilib- network of quartz veinlets is conspicuous in the rium assemblage made up mainly of omphacite (52%) Gorduno lens (Tóth et al. 2000), and assemblages docu- and kyanite (30%) with some phengite and quartz, plus mented from the immediate vicinity of these hydro- trace amounts of garnet, K-feldspar, titanite and rutile fractured zones indicate a relatively high-pressure stage (Table 3). As in domain C, neighboring kyanite and of interaction (ca. 700°C, 8–11 kbar). Yet it appears omphacite porphyroblasts likely produced domain D from the preservation of silica-poor assemblages in sev- upon partial hydration. The two neighboring grains of eral of the domain types that the effect of this hydro- kyanite must have become separated during the early thermal overprint was spatially very limited, possibly stages of symplectite development, after which they due to self-sealing by the positive rVsolid of the horn- evolved as chemically distinct entities. blende-forming reaction. The spatial separation of quartz-undersaturated domains from largely hydrated DISCUSSION matrix containing free quartz is 20–30 m wide, with calcic plagioclase making up most of the armor around Metamorphic differentiation by prograde the relics (Fig. 4). The equilibration volume calculated growth of porphyroblasts for decompression reactions at these conditions is smaller than 104–105 m3. Outside the domains, the The four types of domains documented and modeled homogeneity of the matrix, dominated by hornblende above all represent compositionally distinct reaction and plagioclase, indicates the presence of reaction vol- volumes. If considered in the context of the matrix sur- umes that are at least two or three orders of magnitude rounding these domains, where the amphibolite assem- larger. blage forms a quite homogeneous fabric, the domains The chemical effectiveness of the differentiation es- are unlikely to represent a porphyritic igneous texture. tablished by porphyroblast growth and largely retained Instead, the present assemblages and modes of minerals in the subsequent evolution of the domains is illustrated inside each reaction domain, as well as their shapes, in Figure 5. The local chemical equilibration within pre- point to prograde growth of porphyroblasts as the most viously fractionated chemical domains, with preserva- likely mechanism for domain formation. Lawsonite (and tion of pseudomorphic precursors, such as lawsonite perhaps zoisite), kyanite, and garnet, are all likely to rhomboids in domain-type A, sets limits to the grain- have appeared prior to the complete replacement of pla- scale mass transfer required for the irreversible reactions gioclase by omphacite. The precise sequence of pro- leading to inclusions of Al-rich phases such as corun- grade reactions cannot be deciphered on the basis of the dum, kyanite, and staurolite. Whereas aluminum itself petrographic evidence alone, but comparison with phase is relatively immobile, petrographic evidence in some diagrams computed for the pertinent bulk-compositions cases points to indirect mechanisms for net reaction delimit both the prograde P–T path followed (Fig. 13a) (Carmichael 1969) that assist in reducing local chemi- and the reaction sequence likely experienced by sample cal disequilibrium. The topology of the domains inves- Ma9338. If we assume an early hydration of the igne- tigated in this study (Figs. 4, 8, 10) might indicate that ous basic protolith under (sub?)greenschist-facies con- the most Al-rich product phases only survived in cen- ditions, lawsonite domains (Fig. 7b) then preceded the tral parts of the domains and were replaced within the growth of garnet porphyroblasts (e.g., Figs. 2, 11), and more or less concentric shells armoring the core. This the domains of omphacite (Fig. 12) were established inference requires more careful scrutiny. It may be cor- only after that. Overall, such small-scale metamorphic rect for domain A, where the shell of intermediate pla- differentiation leads to a matrix (~60 vol.%) impover- gioclase (An30–40) is interpreted as a product of reaction ished in Al2O3, CaO, and TiO2 (Fig. 5). derived from an originally larger porphyroblast of lawsonite and the surrounding matrix. For domain B, Reactions during decompression the petrographic evidence and reaction stoichiometries are compatible with a volume-by-volume mechanism of In the evolution following the eclogite-facies stage, replacement of kyanite (plus eclogite matrix phases; the matrix assemblage must have been partially hydrated Table 3) by staurolite (plus plagioclase and minor am- to hornblende, minor plagioclase and <5% quartz. The phibole), if the overall net-transfer reaction (in aqueous fluid gaining access to the relatively small CNFMASH) is written as being fluid-conservative: boudins of eclogite (tens of meters in diameter) is likely 120 THE CANADIAN MINERALOGIST

FIG. 10. Domains C (left in Figs. 9a and 9b) and D (right). (a) Photograph of a thin section in plane-polarized light. (b) Back- scattered SEM image with indexed colors. (c) Variation of plagioclase components (electron-microprobe data) in domain D, along the transect indicated in Figure 9b. (d) Variation of plagioclase components (electron-microprobe data) in domain C, along the transect indicated in Figure 10b. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 121

89 Ky + 235 Cpx + 182 Qtz + 19 Zo + 3 Pg → rated in quartz, in a first stage (20 → 16 kbar, Table 3) 3 St + 276 Pl + 6 Amp (unit: moles) of a sequence of decompression reactions. Also, during that stage, garnet (Alm52Grs25Prp21Sps2) may have Plagioclase is most Al-rich (i.e., calcic) in the inner parts formed at the expense of kyanite and omphacite, of the shell (An78 next to corroded kyanite, An49 next to whereas phlogopite (replacing phengite) and ilmenite newly formed staurolite), and the outermost part of the replacing rutile (partially) were transformations of a plagioclase rim (An26) coexists with Al-free second stage. The sequence by which these phases ap- cummingtonite. Hence, whereas the overall replacement pear in the computed phase-diagram (Fig. 11) is in re- reaction is mass-balanced for composition B’ (Table 3), markably good agreement with the observed texture of the phase topology (Fig. 8) indicates a reaction mecha- the intergrowth in Figure 10, and again the plagioclase- nism that minimized the transport of Al away from the rich shell of the domain can be considered to be part of kyanite precursor. The outer part of the domain is not a the overall domain. rim of metasomatic plagioclase but a reaction product In D, finally, at least two stages of partial re-equili- presumed to be in “closed system” equilibrium with the bration are preserved (Fig. 10). Tiny grains of hercynite core of the domain. We note, however, that the zoning outline what may have been the original boundary of a of the plagioclase indicates a chemical potential gradi- kyanite crystal, and ilmenite completely replaced rutile ent between the outer rim and the core. in the core part of the domain. The newly formed phases For domains C and D, more than one stage of net- are not stable inside the kyanite field, yet sillimanite has transfer reaction is indicated (Fig. 10). In C, the quartz- not been observed. This absence is due either to slow undersaturated assemblage involving corundum may kinetics of transformation of the Al-silicates or a two- have formed early, from an eclogite assemblage satu- stage reaction sequence, with hercynite growing from

FIG. 11. Calculated TKCNFMASH equilibrium phase-diagram delimiting the stability fields of all mineral assemblages for domain C with bulk composition C” (Table 2). Star symbol refers to the predicted stable assemblage listed in Table 3. 122 THE CANADIAN MINERALOGIST

FIG. 12. (a) Calculated TKCNFMASH equilibrium phase-diagram delimiting the stability fields of all mineral assemblages for domain D with bulk composition D” (Table 2). Isopleths for phlogopite in biotite and hercynite in spinel. Star, circle and plus symbols refer to predicted stable assemblages listed in Table 3. (b) Binary join between 0.5 bulk D” and bulk D’, to show the role of quartz. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 123 corundum or staurolite. As in the previous examples (A, starts at the Bøggild gap and ends at the peristerite gap B), the plagioclase mantle around the Al-rich core por- in the phase diagram for albite–anorthite. Overall, we tion is strongly zoned. In this case, the inner portion interpret the distribution and compositional zoning of spans the range from almost pure anorthite (An88) to phases in domain D to be the result of net-transfer reac- An55, the outer portion starts at An38 and is continuously tions in which Al-transport was minimized. The plagio- zoned to An18. This outer part thus shows zoning that clase armor surrounding the partially re-equilibrated central portion shows zoning which we attribute to local equilibration in a chemical gradient. In previous studies, several investigators have put forward hypotheses to explain the presence of minerals like Fe–Mg-bearing spinel, corundum, and sapphirine in kyanite eclogites. Observations of granulite-facies reaction zones 50–100 m across along veins in Nor- wegian eclogites suggest localized fluid-assisted re- equilibration along hydrofractures, leaving the remainder of the eclogite body unchanged (Straume & Austrheim 1999). However, because in our samples hornblende is a dominant phase outside the domains, and the reaction volumes are substantially larger, this model does not explain the textures observed. In other cases, extensive progress of reactions for quartz-undersaturated bulk compositions under granulite-facies conditions led to the formation of sapphirine–plagioclase symplectites after kyanite, but also after garnet and clinopyroxene (Carswell et al. 1989, Godard & Mabit 1998). The observation that domains B, C, and D are surrounded by quartz disqualifies this model for the present case. In addition to the internal texture of the domains studied, the presence of quartz and calcic amphibole in the surrounding matrix renders our symplectites very similar to those described in a kinetic model for the formation of plagioclase coronas around kyanite (Nakamura 2002). In the present study, our thermodynamic modeling with DOMINO and THERIAK, using the estimated bulk compositions of four quite different symplectite domains, proves to be successful in repro- ducing the observed assemblages of minerals. There- fore, the assumptions of local equilibrium and restricted volume of equilibration during symplectite formation, on which the Nakamura model is based in part, are war- ranted.

Decompression paths inferred for lenses inside the tectonic accretion channel

The models presented above for four domains in kyanite eclogites from the Central Alps yield new constraints on the P–T trajectories of the boudins from which they were collected. The mineral assemblage in domain A (Loc. 1, Fig. 1b) yields conditions of equilibration of 675 ± 25°C and 6.5 ± 1 kbar. This sample FIG. 13. (a) Estimates of P–T evolution, domain A; P–T tra- was also included in the analysis of Tóth et al. (2000), jectory in grey from Tóth et al. (2000) for this same lens. and the present results provide additional constraints on The prograde path is now constrained to blueschist-facies conditions, owing to evidence of a lawsonite precursor in the P–T trajectory derived in that paper (Fig. 13). The domain A. (b) P–T estimates for the partial equilibration of rocks were shown to have undergone high-pressure domains B, C and D, from another boudin of eclogite in the metamorphism at 23 ± 3 kbar and 750 ± 50°C, subse- same mélange zone. Note the different scales of tempera- quently cooled during rapid decompression to 675 ± ture in a and b. 25°C at 8 ± 1 kbar, before a final heating phase at simi- 124 THE CANADIAN MINERALOGIST lar pressures and 750 ± 40°C. The present result sug- scattered electron images yields accurate estimates of gests that initial cooling proceeded to somewhat lower the surface area of each phase present. The assumption pressures. Importantly, our analysis indicates that that surface area in thin sections can be directly con- lawsonite is the most likely precursor for the domain, verted to modal volume is apparently valid for the kind implying that the subduction took place in a much cooler of texture studied. The choice of an average composi- setting than previously considered (Fig. 4b). tion for each phase which, when combined with their Domains B, C and D stem from a single boudin (Loc. modal volume, yields an estimate for the chemical com- 2, Fig. 1b), and although they have recorded somewhat position of the domains, introduces additional uncer- different conditions of equilibration, they share the same tainty. Electron-microprobe data for the individual P–T trajectory. The estimated conditions of equilibra- phases, an assessment of the variety of each mineral, tion for the three domains constrain the Barrovian meta- and zoning patterns support an appropriate choice, morphic stage, as experienced by these samples (Fig. which is in some cases hampered by the small size of 13). The arrow indicates a trajectory that the boudin is grains in the symplectites. Finally, the amount and com- likely to have followed, although the temperature re- position of fluids present at the time of mineral reaction corded by domain C may actually have been up to 100°C (or re-equilibration) are difficult to assess, although in higher. High-pressure metamorphism and initial decom- some cases, an identified precursor phase, such as pression thus were followed by a stage of near-isobaric lawsonite in domain A, imposes limits (Fig. 7a). In other heating to about 750°C, which has also been reported cases, a variety of preliminary scenarios may be formu- previously from other bodies of eclogite in the area (e.g., lated, depending on the availability of fluids, which Brouwer 2000, Tóth et al. 2000). An investigation of possibilities can then be assessed by calculating binary other domain types from this eclogite lens may yield phase-diagrams, with variable amounts of the volatile further constraints on the P–T conditions during sub- component present, as a function of pressure or tem- duction, high-pressure metamorphism and initial de- perature. compression in the mélange sequence from which the samples were taken. Owing to its different chemical CONCLUSIONS composition, these conditions did not lead to the stabi- lization of hercynite-dominant spinel or corundum in Kyanite eclogites develop plagioclase symplectites domain B (Fig. 8). In any case, the P–T conditions for involving high-Al phases like corundum, Fe–Mg-bear- the Barrovian overprint recorded in this lens are above ing spinel and, in rare cases, staurolite, which would not 700°C between about 9 and 4 kbar. This temperature is normally be expected in rocks of mafic bulk-composi- only slightly higher than that recorded in the same area tion. Thermodynamic modeling of such domains in for the regional Tertiary overprint, for pressure condi- eclogites from the Central Alps indicates that during tions of 5.5–6 kbar (Todd & Engi 1997). decompression, such phases can develop where the equilibration volume involving porphyroblasts is re- Uncertainties in the approach stricted, rendering the effective bulk-composition dis- tinctly different from that of a normal MORB-type A major uncertainty involved in thermodynamic basalt. Accurate estimates of domain composition are modeling of mineral assemblages is related to the ther- obtained by combining image analysis with electron- modynamic database used. The current weakness of microprobe data, and may be refined in an iterative pro- solution models to adequately describe complex am- cess during which mineral assemblages predicted by phiboles, and the limited applicability of solution mod- thermodynamic models are compared with observa- els for clinopyroxene at temperatures below about tions. The fact that even models based on initial esti- 500°C limit the types of assemblages that can be mod- mates of domain composition reproduce most features eled. In the four domains presented here, amphibole is of the domains studied supports this approach, and in- present in very small amounts only (<2 vol.%). The dicates that the basic assumption of local equilibrium is pargasite – actinolite solution model adopted here yields probably tenable. Once the domains are accurately only an approximate description of the magnesio- documented, thermodynamic models are used to ana- hornblende present, but owing to the low modal abun- lyze the conditions at which the symplectites developed, dance, this inaccuracy should not affect the other phases and in some cases precursor phases can be inferred. In very strongly. We note, however, that isopleths for the three samples analyzed in detail, lawsonite and kyanite composition of amphiboles cannot be used to infer ex- porphyroblasts formed at the prograde stage. Local tex- act conditions of equilibration. tures produced on the decompression path and quanti- A second uncertainty lies in the determination of the tative phase-relations for that stage indicate small “bulk” compositions of domains. The models presented reaction-volumes and minimum distances of transport above closely approximate the observed assemblages of of Al-carrier species. minerals, which suggests that determining the domain The results of this analysis, applied to kyanite boundaries optically, and iterative adjustment based on eclogite samples from the Southern Steep Belt of the model results, have been effective. Analysis of back- Central Swiss Alps, provides new constraints on pro- DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 125 grade and retrograde P–T paths, and they confirm late- Gaggio-Basal (cantone Ticino). Schweiz. Mineral. Petrogr. stage heating at mid-crust depths. Mitt. 33, 173-480.

DE CAPITANI, C. (1994): Gleichgewichts-Phasendiagramme: ACKNOWLEDGEMENTS theorie und Software. Eur. J. Mineral. 6 Beiheft, 48 [http:/ /titan.minpet.unibas.ch/minpet/groups/theriak/theruser. It is our pleasure to acknowledge particularly per- html]. ceptive comments on this manuscript from Dugald M. Carmichael, to whom we dedicate this paper. He has ELVEVOLD, S. & GILOTTI, J.A. (2000): Pressure–temperature guided us and a generation of petrologists by his stud- evolution of retrogressed kyanite eclogites, Weinschenk ies, combining astute observation, eloquent descriptions, Island, North-East Greenland Caledonides. Lithos 53, 127- and rigorous thinking, and we have all been blessed with 147. his joyful personality. Discussions with Alfons Berger and Tom Burri, and ENAMI, M. & ZANG, Q. (1988): Magnesian staurolite in gar- analytical assistance from Nicola Bistacchi have con- net–corundum rocks and eclogite from the Donghai dis- trict, Jiangsu province, east China. Am. Mineral. 73, 48-56. tributed to the substance of this study. Thoughtful re- views by Marc R. St-Onge and Encarnación Puga, as ENGI, M. (1983): Equilibria involving Al–Cr spinel. I. Mg–Fe well as careful editorial comments from Robert F. Mar- exchange with olivine; experiments, thermodynamic analy- tin, have helped us improve the presentation. Financial sis, and consequences for geothermometry. Am. J. Sci. 283- support from Schweizerischer Nationalfonds (No. 20– A, 29-71. 63593.00 and 20020–101826) is gratefully acknowl- edged. ______, BERGER, A. & ROSELLE, G.T. (2001): Role of the tectonic accretion channel in collisional orogeny. Geology 29, 1143-1146. REFERENCES

GHIORSO, M.S., EVANS, B.W., HIRSCHMANN, M.M. & YANG, BÄCHLIN, R., BIANCONI, F., CODONI, A., DAL VESCO, E., HEXIONG (1995): Thermodynamics of the amphiboles; Fe– KNOBLAUCH, P., KÜNDIG, E., REINHARD, M., SPAENHAUER, Mg cummingtonite solid solution. Am Mineral. 80, 502- F., SPICHER, A., TROMMSDORFF, V. & WENK, E. (1974): 519. Geologischer Atlas der Schweiz, 1:25 000, Blatt 1313 Bellinzona. Schweizerische Geologische Kommission, GODARD, G. & MABIT, J.-L. (1998): Peraluminous sapphirine Bern, Switzerland. formed during retrogression of a kyanite-bearing eclogite from Pays de Léon, Armorican Massif, France. Lithos 43, BERMAN, R.G. (1988): Internally-consistent thermodynamic 15-29. data for minerals in the system Na2O–K2O–CaO–MgO– FeO–Fe O –Al O –SiO –TiO –H O–CO . J. Petrol. 29, 2 3 2 3 2 2 2 2 GRUBENMANN, U. (1908): Granatolivinfels des Gordunotales 445-522. und seine Begeleitgesteine. Viertelj. Naturf. Ges. Zürich 53, 129-156. BOCCHIO, R., LIBORIO, G. & MOTTANA, A. (1985): Petrology of the amphibolitized eclogites of Gorduno, Lepontine Alps, HEINRICH, C.A. (1986): Eclogite facies regional metamorphism Switzerland. Chem. Geol. 50, 65-86. of hydrous mafic rocks in the Central Alpine Adula nappe. J. Petrol. 27, 123-154. BROUWER, F.M. (2000): Thermal Evolution of High-Pressure Metamorphic Rocks in the Alps. Ph.D. thesis, Utrecht Uni- JOHANSSON, L. & MÖLLER, C. (1986): Formation of sapphirine versity. Geologica Ultraiectina 199, 221 p. during retrogression of a basic high-pressure granulite, Western Gneiss region, Norway. Contrib. Mineral. Petrol. CARMICHAEL, D.M. (1969): On the mechanism of prograde 94, 29-41. metamorphic reactions in quartz-bearing pelitic rocks. Contrib. Mineral. Petrol. 20, 244-267. KRETZ, R. (1983): Symbols for rock-forming minerals. Am. Mineral. 68, 277-279. CARPENTER, M.A. (1985): Order–disorder transformations in mineral solid solutions. In Microscopic to Macroscopic: LAPPIN, M.A. (1960): On the occurrence of kyanite in the Atomic Environments to Mineral Thermodynamics (S.W. eclogites of the Selje and Åheim districts, Nordfjord. Nor. Kiefer & A. Navrotsky, eds.). Rev. Mineral. 14, 187-223. Geol. Tidsskr. 40, 289-296.

CARSWELL, D.A., MÖLLER, C. & O’BRIEN, P.J. (1989): Origin LIATI, A. & SEIDEL, E. (1994): Sapphirine and högbomite in of sapphirine–plagioclase symplectites in metabasites from overprinted kyanite-eclogites of central, N. Greece: first Mitterbachgraben, Dunkelsteinerwald granulite complex, evidence of granulite-facies metamorphism. Eur. J. Min- Lower Austria. Eur. J. Mineral. 1, 455-466. eral. 6, 733-738.

DAL VESCO, E. (1953): Genesi e metamorfosi delle rocce basiche MÄDER, U.K. & BERMAN, R.G. (1992): Amphibole thermo- e ultrabasiche nell’ambiente mesozonale dell’orogene barometry: a thermodynamic approach. Geol. Surv. Can., pennidico: studio geologico-petrografico della catena Pap. 92–1E, 393-400. 126 THE CANADIAN MINERALOGIST

MEYRE, C., DE CAPITANI, C. & PARTZSCH, J.H. (1997): A ter- from the South Carpathians: status and implications. Int. nary solid solution model for omphacite and its application Geol. Rev. 44, 859-876. to geothermobarometry of eclogites from the Middle Adula nappe (Central Alps, Switzerland). J. Metamorph. Geol. 15, SCHMID, S.M., AEBLI, H.R., HELLER, F. & ZINGG, A. (1989): 687-700. The role of the Periadriatic Line in the tectonic evolution of the Alps. In Alpine Tectonics (M.P. Crawford, D. ______, ______, ZACK, T. & FREY, M. (1999): Petrology Dietrich & R.G. Park, eds.). Geol. Soc., Spec. Publ. 45, of high-pressure metapelites from the Adula nappe (Cen- 153-171. tral Alps, Switzerland). J. Petrol. 40, 199-213. ______, PFIFFNER, O.A., FROITZHEIM, N., SCHÖNBORN, G. & MÖCKEL, J.R. (1969): Structural petrology of the garnet- KISSLING, E. (1996): Geophysical-geological transect and peridotite of Alpe Arami (Ticino, Switzerland). Leidse tectonic evolution of the Swiss–Italian Alps. Tectonics 15, Geol. Meded. 42, 61-130. 1036-1064.

NAGEL, T. (2002): Metamorphic and Structural History of the STRAUME, Å.K. & AUSTRHEIM, H. (1999): Importance of frac- Southern Adula Nappe (Graubünden, Switzerland). Ph.D turing during retro-metamorphism of eclogites. J. thesis, Geologisch-Paläontologischen Institut, University Metamorph. Geol. 17, 637-652. of Basel, Basel, Switzerland. TODD, C.S. & ENGI, M. (1997): Metamorphic field gradients in NAKAMURA, D. (2002): Kinetics of decompressional reactions the Central Alps. J. Metamorph. Geol. 15, 513-530. in eclogitic rocks – formation of plagioclase coronas around kyanite. J. Metamorph. Geol. 20, 325-333. TÓTH, T.M., GRANDJEAN, V. & ENGI, M. (2000): Polyphase evolution and reaction sequence of compositional domains PFIFFNER, M. & TROMMSDROFF, V. (1998): The high-pressure in metabasalt: a model based on local chemical equilibrium ultramafic-mafic-carbonate suite of Cima Lunga–Adula, and metamorphic differentiation. Geol. J. 35, 163-183. Central Alps: excursions to Cima di Gagnone and Alpe Arami. Schweiz. Mineral. Petrog. Mitt. 78, 337-354.

SABAU, G., ALBERICO, A. & NEGELESCU, E. (2002): Peralu- Received October 22, 2003, revised manuscript accepted minous sapphirine in retrogressed kyanite-bearing eclogites June 22, 2004. DOMAIN EVOLUTION IN ECLOGITE FROM THE CENTRAL SWISS ALPS 127 128 THE CANADIAN MINERALOGIST